Understanding Lactate in an Intensive Care Setting
by
Hilary G. Mulholland
S.B., Massachusetts Institute of Technology (2014)
Submitted to the Department of Electrical Engineering and Computer
Science
in partial fulfillment of the requirements for the degree of
Master of Engineering in Computer Science and Molecular Biology
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2015
c Massachusetts Institute of Technology 2015. All rights reserved.
○
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Electrical Engineering and Computer Science
May 18, 2015
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Roger G. Mark
Distinguished Professor in Health Sciences and Technology and
Electrical Engineering and Computer Science
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Prof. Albert R. Meyer
Chairman, Masters of Engineering Thesis Committee
2
Understanding Lactate in an Intensive Care Setting
by
Hilary G. Mulholland
Submitted to the Department of Electrical Engineering and Computer Science
on May 18, 2015, in partial fulfillment of the
requirements for the degree of
Master of Engineering in Computer Science and Molecular Biology
Abstract
We investigated the relationship between initial lactate levels and ICU patient outcomes using the MIMIC II (version 2.6) database. We divided ICU admissions based
on their initial lactate measurement into three groups: admissions with high lactate
(above 4.0 mmol/L), admissions with medium lactate (between 2.0 mmol/L and 4.0
mmol/L), and admissions with low lactate (below 2.0 mmol/L). In addition to the
ICU population as a whole, we studied sepsis patients using three different criteria
(Martin, Angus, and infection with SIRS).
We found that increased lactate levels were associated with a higher ICU mortality,
higher 30 day mortality, longer ICU length of stay, and higher SOFA and SAPS I
severity scores in all ICU admissions and in all three sepsis cohorts. Sepsis patients
with high initial lactate levels were the most severely ill of all the patient populations.
Sepsis patients identified with the Martin criteria who had high lactate levels had the
worst outcomes of the three sepsis cohorts, but had similar average severity scores.
This suggests that knowing lactate levels may give predictive value in addition to
severity scores.
We also investigated the relationship between initial lactate, change in lactate
from the first measurement to the second measurement, and ICU mortality. We
found that patients with high initial lactate levels in combination with an increase
in lactate level typically had poorer outcomes than patients with high initial lactate
levels with a decrease in lactate level.
Thesis Supervisor: Roger G. Mark
Title: Distinguished Professor in Health Sciences and Technology and Electrical Engineering and Computer Science
3
4
Acknowledgments
First and foremost, I would like to thank my thesis advisor Professor Roger Mark.
Without his guidance and advice, my thesis would not have been possible. He always
took the time to meet with me, and he helped me keep progressing throughout my
time at the LCP.
I would also like to thank Ikaro Silva. He first introduced me to the project, and
continued to work with me through the rest of my thesis. He answered any questions
I asked, and he gave me interesting suggestions to pursue throughout the project. He
came to every meeting, and he was invested in the project and my success which I
really appreciated.
In addition, I would like to thank the rest of the LCP. Dr. Leo Celi gave us
clinical advice and provided motivation for the project. I loved attending the lab’s
group meetings, and hearing about the research everyone in the lab was doing.
Finally, I would like to thank my parents and the rest of my family and friends for
supporting me. I especially would like to thank my grandmother Geraldine Kempton
for her generosity and support.
I have learned so much this past year, and I am very grateful for all of the guidance
and help I received along the way.
5
6
Contents
1 Introduction
15
1.1
Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
1.2
Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
2 Background
17
2.1
Lactate Metabolism . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.2
Hyperlactatemia
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
2.2.1
Causes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
2.2.2
Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
Role of Lactate in Emergency Medicine . . . . . . . . . . . . . . . . .
21
2.3.1
Sepsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
Thesis Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
2.3
2.4
3 Methods
27
3.1
Cohort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
3.2
Initial Lactate Measurements . . . . . . . . . . . . . . . . . . . . . .
28
3.3
Parameters Compared to Lactate . . . . . . . . . . . . . . . . . . . .
29
3.3.1
ICU Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
3.3.2
30 Day Mortality . . . . . . . . . . . . . . . . . . . . . . . . .
29
3.3.3
Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . .
29
3.3.4
Severity of Illness . . . . . . . . . . . . . . . . . . . . . . . . .
29
3.4
Conditions Associated with High Lactate . . . . . . . . . . . . . . . .
30
3.5
Change in Lactate . . . . . . . . . . . . . . . . . . . . . . . . . . . .
30
7
3.6
Septic Cohort . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.6.1
Martin Implementation . . . . . . . . . . . . . . . . . . . . . .
31
3.6.2
Angus Implementation . . . . . . . . . . . . . . . . . . . . . .
31
3.6.3
Infection and SIRS . . . . . . . . . . . . . . . . . . . . . . . .
32
4 Lactate Associations in All ICU Patients
4.1
4.2
4.3
Associations with Initial Lactate Levels . . . . . . . . . . . . . . . . .
33
4.1.1
ICU Mortality Rate . . . . . . . . . . . . . . . . . . . . . . . .
33
4.1.2
30 Day Mortality Rate . . . . . . . . . . . . . . . . . . . . . .
34
4.1.3
Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . .
35
4.1.4
Severity of Illness . . . . . . . . . . . . . . . . . . . . . . . . .
36
4.1.5
Change in Lactate . . . . . . . . . . . . . . . . . . . . . . . .
37
Lactate Associations in Different Care Units . . . . . . . . . . . . . .
40
4.2.1
ICU Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
4.2.2
30 Day Mortality . . . . . . . . . . . . . . . . . . . . . . . . .
41
4.2.3
Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . .
42
Conditions Associated with High Lactate . . . . . . . . . . . . . . . .
42
5 Lactate Associations in Septic Patients
5.1
5.2
33
45
Martin Criteria Severe Sepsis . . . . . . . . . . . . . . . . . . . . . .
45
5.1.1
ICU Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . .
45
5.1.2
30 Day Mortality . . . . . . . . . . . . . . . . . . . . . . . . .
46
5.1.3
Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . .
47
5.1.4
Severity of Illness . . . . . . . . . . . . . . . . . . . . . . . . .
48
5.1.5
Change in Lactate . . . . . . . . . . . . . . . . . . . . . . . .
49
Angus Criteria Severe Sepsis . . . . . . . . . . . . . . . . . . . . . . .
50
5.2.1
ICU Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
5.2.2
30 Day Mortality . . . . . . . . . . . . . . . . . . . . . . . . .
51
5.2.3
Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . .
52
5.2.4
Severity of Illness . . . . . . . . . . . . . . . . . . . . . . . . .
53
5.2.5
Change in Lactate . . . . . . . . . . . . . . . . . . . . . . . .
54
8
5.3
Infection and SIRS Criteria Sepsis . . . . . . . . . . . . . . . . . . . .
55
5.3.1
ICU Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . .
55
5.3.2
30 Day Mortality . . . . . . . . . . . . . . . . . . . . . . . . .
56
5.3.3
Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . .
57
5.3.4
Severity of Illness . . . . . . . . . . . . . . . . . . . . . . . . .
58
5.3.5
Change in Lactate . . . . . . . . . . . . . . . . . . . . . . . .
59
6 Discussion
61
6.1
All ICU Admissions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
6.2
Septic Admissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
6.3
Change in Lactate . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
6.4
Comparisons Across Cohorts . . . . . . . . . . . . . . . . . . . . . . .
64
7 Conclusion
67
7.1
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
7.2
Future Research Directions . . . . . . . . . . . . . . . . . . . . . . . .
68
A
71
A.1 Antibiotics used in Sepsis Definition . . . . . . . . . . . . . . . . . . .
9
71
10
List of Figures
2-1 Lactate production from pyruvate . . . . . . . . . . . . . . . . . . . .
18
4-1 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Below 2.0 mmol/L . .
38
4-2 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Between 2.0 mmol/L and
4.0 mmol/L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
4-3 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L . .
39
4-4 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality for All ICU Admissions with Any Lactate . . . . . . . . . .
40
5-1 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L for
Sepsis Admissions Defined by Martin Criteria . . . . . . . . . . . . .
50
5-2 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L for
Sepsis Patients Defined by Angus Criteria . . . . . . . . . . . . . . .
55
5-3 Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L for
Sepsis Admissions Defined by Infection and SIRS Criteria . . . . . . .
11
60
12
List of Tables
2.1
Causes of Hyperlactatemia . . . . . . . . . . . . . . . . . . . . . . . .
20
2.2
Criteria for SIRS, Sepsis, Severe Sepsis, and Septic Shock . . . . . . .
23
3.1
Cohort broken down into groups by lactate measurement . . . . . . .
28
4.1
ICU Mortality Vs. Lactate for All ICU Admissions . . . . . . . . . .
34
4.2
30 Day Mortality Vs. Lactate for All ICU Admissions . . . . . . . . .
35
4.3
Average and Median Length of Stay Vs. Lactate for All ICU Admissions Who Survived Their ICU Stay . . . . . . . . . . . . . . . . . . .
36
4.4
Average SOFA and SAPS I Scores Vs. Lactate for All ICU Admissions
37
4.5
ICU Mortality of Admissions With a Lactate Measurement in Each
ICU Care Unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6
30 Day Mortality of Admissions With a Lactate Measurement in Each
ICU Care Unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.7
42
Average and Median Length of Stay of Admissions With Lactate Levels
Broken Down By Care Unit . . . . . . . . . . . . . . . . . . . . . . .
4.8
41
43
Primary ICD-9 Codes Most Commonly Associated with High Lactate
Levels (Of a Total of 2,240 Admissions With High Lactate Levels) . .
44
5.1
ICU Mortality Vs. Lactate for Martin Criteria Severe Sepsis Patients
46
5.2
30-Day Mortality Vs. Lactate for Martin Criteria Severe Sepsis Patients 47
5.3
Average and Median Length of Stay for Sepsis Admissions Defined by
Martin Criteria Who Survived Their ICU Stay . . . . . . . . . . . . .
13
48
5.4
Average SOFA and SAPS I Scores Sepsis Admissions Defined by Martin
Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5
ICU Mortality Vs. Lactate for Sepsis Admissions Defined by Angus
Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.6
. . . .
53
Average SOFA and SAPS I Scores Vs. Lactate for Sepsis Admissions
Defined by Angus Criteria . . . . . . . . . . . . . . . . . . . . . . . .
5.9
52
Average and Median Length of Stay Vs. Lactate for Sepsis Admissions
Defined by the Angus Criteria Who Survived Their ICU Stay
5.8
51
30-Day Mortality Vs. Lactate for Sepsis Admissions Defined by Angus
Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7
49
54
ICU Mortality Vs. Lactate for Sepsis Patients Defined by Infection and
SIRS Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
5.10 30 Day Mortality Vs. Lactate for Sepsis Patients Defined by Infection
and SIRS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
5.11 Average and Median Length of Stay for Sepsis Admissions Defined by
Infection and SIRS Who Survived Their ICU Stay . . . . . . . . . . .
58
5.12 Average SOFA and SAPS I Scores Vs. Lactate for Sepsis Admissions
Defined by Infection and SIRS Criteria . . . . . . . . . . . . . . . . .
14
59
Chapter 1
Introduction
1.1
Motivation
Lactate is a molecule continuously created in the body that has been used as a marker
for declining health of intensive care unit (ICU) patients. Lactate has often been used
as a surrogate for strained cellular metabolism and may have strong predictive value
in determining outcome.
Clinicians consider persistent high lactate levels to be a sign of poor health. There
have been many previous research studies focused on the predictive value of lactate
clearance rates in ICU patients. Lactate levels can change very rapidly, so measurements need to be taken regularly in order to stay accurate. However, lactate cannot
be measured without drawing arterial or venous blood which is a fairly invasive procedure that can lead to in hospital infections. Because of this risk, doctors may not
order lactate levels to be measured as frequently, and they can miss rapid increases
in lactate [1].
There have been few studies that focus on the predictive value of the initial lactate
measurement in ICU patients. We believed that much could be learned about a
patient’s illness by knowing the initial lactate level, rather than waiting to see the
change in lactate. In addition to lactate clearance rates, we were interested to see the
association between the initial lactate level and patient outcome.
Gaining a better understanding of initial lactate measurements in ICU patients
15
may be useful to clinicians so that they may be able to predict outcome more effectively and be able to make better treatment plans. Quantifying the associations
between initial lactate level and patient outcome may give insight as to how lactate
levels can be used to diagnose illness in the ICU.
1.2
Thesis Outline
This thesis contains 7 chapters and one appendix. A brief overview of each chapter
is as follows:
∙ Chapter 2: Background. This chapter reviews lactate metabolism and previous
studies focused on lactate levels in ICU patients.
∙ Chapter 3: Methods. This chapter details the cohort we studied and the methods we followed to quantify relationships with lactate.
∙ Chapter 4: Lactate Associations in All ICU Patients. This chapter reviews all
of the results we obtained when following our methods using a cohort of ICU
patient.
∙ Chapter 5: Lactate Associations in Septic Patients. This chapter reviews all
of the results we obtained when following our methods using cohorts sepsis
patients in the ICU.
∙ Chapter 6: Discussion. This chapter analyzes and discusses the results found
in Chapter 4 and 5.
∙ Chapter 7: Conclusions and Future Research Directions. This chapter summarizes this thesis and gives further research directions.
16
Chapter 2
Background
Accumulation of lactate in the body has been associated with poor outcomes in
clinical settings. As the level of lactate increases above the normal range, the consequences increase in severity and outcomes are typically worse. Understanding lactate
metabolism and its implications for the physiologic function of the body is very important for a patient’s diagnosis and outcome.
2.1
Lactate Metabolism
Most cell energy is created in the form of ATP aerobically through the citric acid cycle
and the electron transport chain. This process can only take place in environments
with sufficient tissue resources and oxygen [2]. Pyruvate, created during glycolysis,
is converted to acetyl CoA during the citric acid cycle, creating NADH. The overall
reaction is as follows:
pyruvate + NAD+ + CoA ←−→ Acetyl CoA + CO2 + NADH
NADH is then sent to the electron transport chain where ATP is created through
oxidative phosphorylation. In total through aerobic respiration, one glucose molecule
creates 36 ATPs through glycolysis, the citric acid cycle and the electron transport
chain.
17
If there is inadequate tissue perfusion or a lack of oxygen, the pyruvate created
during glycolysis is instead converted anaerobically to lactate. Lactate dehydrogenase
catalyzes the conversion between pyruvate and lactate as shown in Figure 1-1 [1].
NADH and a proton are consumed during this reaction. Only 2 ATP are created
through substrate-level phosphorylation during glycolysis compared to the 36 ATP
extracted during aerobic respiration.
Figure 2-1: Lactate production from pyruvate
For every glucose molecule that goes through glycolysis and then undergoes anaerobic respiration, 2 lactate molecules are produced along with 2 protons and 2 ATPs.
The overall reaction is as follows:
glucose + 2 (ADP + inorganic phosphate) ←−→ 2 lactate + 2 H+ + 2 ATP
Even when oxygen is present, lactate is kept at a steady basal level in many tissues
including skeletal muscle, the brain, kidneys, and red blood cells [2]. Daily, the human
body creates approximately 20 mmol of lactate per kilogram of body weight. Lactate
is also consumed at a high rate of about 320 mmol per liter per hour to keep levels
low. Clearance of the created lactate keeps the level of lactate in the blood stable
below 1 mmol per liter in normal conditions. Lactate is converted back to pyruvate
through various processes such as gluconeogenesis in the liver and kidney and also the
tricarboxylic acid cycle and oxidative phosphorylation in various organs throughout
the body. Overall the liver is responsible for 70% of lactate clearance in the body.
These processes keep the ratio of pyruvate to blood lactate at 10:1. Lactate is kept
18
constant in all tissues by lactate shuttles that help moderate the movement of lactate
throughout the body. By these clearance processes, protons are also consumed which
helps maintain acid-base homeostasis throughout the body.
2.2
Hyperlactatemia
Hyperlactatemia occurs when the production of lactate is greater than the consumption of lactate or clearance of lactate is malfunctioning. Hyperlactatemia implies addition of protons equal to the number of excess lactate molecules. The extra protons
may cause an acid-base imbalance, so hyperlactatemia is also called lactic acidosis.
2.2.1
Causes
In normal situations, lactate can increase greatly but usually is consumed equally as
quickly. During intense exercise, lactate concentration can increase by a factor of 100,
but it quickly decreases back to normal levels after exercise. Shivering and seizures
can also cause acute lactate increases, but they are normally cleared quickly.
Various other disorders account for more serious lactic acidosis. Lactic acidosis
can be divided into two types. Lactic acidosis is type A if its cause is associated with
tissue hypoxia, and it is type B if there is no tissue hypoxia [2]. A patient may have
both types if multiple causes are contributing to the overproduction of lactate. The
most common causes of hyperlactatemia are cardiogenic shock, severe heart failure,
trauma, and sepsis.
Causes of increased lactate production can further be divided into three main categories: Demand, Delivery, and Utilization [1]. Increased demand for oxygen, inadequate oxygen delivery to tissues, and inadequate utilization of oxygen all contribute
to the production of lactate. Table 1.1 shows specific disorders corresponding to each
cause of lactic acidosis. Drugs affecting utilization that may cause lactic acidosis
include Metformin, nucleoside reverse-transcriptase inhibitors, and propofols. These
drugs interfere with oxidative phosphorylation causing lactate to be produced anaerobically. Thiamine deficiency causes reduction of pyruvate dehydrogenase activity,
19
Inadequate
Delivery of 𝑂2
Increased
Demand for 𝑂2
Inadequate
Utilization of 𝑂2
Volume depletion
Significant blood loss
Septic shock
Profound anemia
Severe hypoxemia
Carbon monoxide
exposure
Trauma
Shivering
Seizures
Strenuous exercise
SIRS
Diabetes mellitus
Thiamine Deficiency
HIV infection
Certain drugs
Table 2.1: Causes of Hyperlactatemia
thus increasing lactate production.
There are a few causes that also do not fit into any of the three categories mentioned. Cancer can cause an increase in glycolysis in a tumor, and the excess product
is converted to lactate. In liver disease and severe liver cancer, the clearance of lactate
decreases which overall increases the amount in the body. Multiple disorders may act
at the same time to increase lactate levels even more significantly.
2.2.2
Treatment
Hyperlactatemia has been associated with poor outcomes in clinical settings, so monitoring high lactate levels is a main priority. Normal blood lactate levels range from
approximately 0.5 mmol per liter to an upper limit of 2.0 mmol per liter. Lactate in
a patient can be measured either in venous or arterial blood. Knowing the underlying cause of the increased lactate concentration is very important in determining
treatment [2].
When oxygen delivery is the main cause of lactic acidosis, restoring tissue perfusion
is very important. Oxygen delivery can depend on cardiac output, hemoglobin, or the
partial pressure of oxygen in the blood. Vasopressors and other agents are often used
to help restore perfusion in cases of low blood pressure. In situations where sepsis is
the main cause, antibiotics are given to help treat the underlying infection causing
20
the lactate increase. Treating the main cause is often how increased lactate levels are
addressed in a clinical setting.
2.3
Role of Lactate in Emergency Medicine
In the emergency department, knowing a patient’s lactate levels may provide useful
information for treatment and predicting outcome. Studies have shown that an increased lactate level is associated with increased mortality in patients that have sepsis,
and the higher the lactate level, the worse the outcome will be. In addition, studies
have been done to show that if lactate is cleared within a short period of time (about
24 hours), the outcome is much more likely to be positive. Lactate measurements
provide physicians with information about the patient’s cellular metabolism and the
overall severity of their illness especially in sepsis or trauma patients.
In situations of severe injury, lactate levels give information about how cellular
metabolism is affected and can be used to predict outcome. In 1993, Abramson et
al. performed a study looking at the survival rates of 76 trauma patients admitted
into the Intensive Care Unit (ICU) [3]. All of the 27 patients whose lactate cleared
to a normal level below 2.0 mmol per liter within 24 hours survived. When lactate
was cleared between 24 and 48 hours, only 75% survived, and when lactate was not
cleared within 48 hours only 14% survived (3 out of 22). They concluded that lactate
clearance time was very important in determining patient outcomes, while initial
lactate measurement did not seem as helpful.
Shapiro et al. found a connection between lactate level and mortality when studying 1,278 patients with infections [4]. Patients with a lactate level less than 2.5 mmol
per liter had a mortality rate of only 4.9% whereas patients with lactate over 4 mmol
per liter had a mortality rate of 28.4%. As lactate levels increased, the outcomes
became worse.
21
2.3.1
Sepsis
In the United States, severe sepsis has become a very prevalent syndrome that has
become a large focus in medicine. In 2001, Angus et al. concluded that there were
over 750,000 cases of severe sepsis in the United States each year, and since then the
prevalence has only increased [5]. Severe sepsis patients account for 2% of all patients
that are admitted to the hospital and of those patients half of them are admitted into
the ICU. Severe sepsis patients represent 10% of all ICU admissions. Understanding
sepsis and its causes is important for improving outcomes and diagnosis.
Sepsis, Severe Sepsis, and Septic Shock
Sepsis is defined as a systemic inflammatory response to infection. For a patient to be
considered septic, they must have a documented or suspected infection and have signs
of systematic inflammatory response syndrome (SIRS) [6]. SIRS is diagnosed when
a patient has two or more of the following symptoms: body temperature less than
36∘ C or greater than 38∘ C, heart rate above 90 beats per minute, respiratory rate
higher than 20 per minute or partial pressure of carbon dioxide lower than 32 mmHg,
white blood cell count higher than 12,000 cells per microliter or lower than 4,000
cells per microliter or greater than 10% immature (band) forms. The most common
cause of infection in septic patients is pneumonia which is associated with half of all
cases. Intrabdominal and urinary tract infections are the next most common causes.
Infections causing sepsis can be acquired in the community and also in health care
facilities.
Sepsis can become very critical if it is not treated quickly; patients may become
severely septic or go into septic shock. Severe sepsis is defined as sepsis complicated
by organ dysfunction, and septic shock is defined as sepsis with persistent arterial
hypotension. Hypotension is defined as systolic arterial pressure below 90 mmHg,
mean arterial pressure lower than 60 mmHg, or a reduction in systolic blood pressure
of more than 40 mmHg. Persistent hypotension refers to hypotension that persists
even after volume resuscitation [7]. Table 1.2 summarizes the criteria of sepsis, severe
22
sepsis, and septic shock.
Table 2.2: Criteria for SIRS, Sepsis, Severe Sepsis, and Septic Shock
Term
Criteria
SIRS
At least 2 of the following:
∙ Temperature > 38∘ C or < 36∘ C
∙ Heart rate > 90/min
∙ Hyperventilation evidenced by respiratory rate >20/min or
arterial 𝐶𝑂2 lower than 32 mmHg
∙ White blood cell count >12000 cells/𝜇L or lower than 4000
cells/𝜇L or >10% immature band forms
Sepsis
SIRS criteria with suspected or proven infection
Severe Sepsis
Sepsis with organ dysfunction
Septic Shock
Sepsis with hypotension despite adequate fluid resuscitation
Lactate in Sepsis Patients
Studies have been done to show the importance of lactate in diagnosing and treating
severe sepsis and septic shock. In an analysis of 20 hemodynamic variables and organ
dysfunction variables, lactate was found to be the only easily determined parameter
that was helpful in the prediction of outcome in septic patients [8]. Lactate has been
found to be an important parameter for indicating sepsis induced hypoperfusion and
also an important indicator for early goal directed therapy of sepsis. An increase in
lactate in septic patients may not be due to decreased perfusion though, and it may
be due to cellular dysfunction resulting from toxins instead.
The rate of clearance of lactate has been associated with the outcomes of patients.
Nguyen et al. studied the lactate clearance of 111 ICU patients that had either severe
sepsis or septic shock. Lactate clearance was defined as the percent decrease in lactate
23
after 6 hours. They found that as lactate clearance increased, mortality decreased.
For approximately every 10% increase in lactate clearance, there was an 11% decrease
in mortality rate. In contrast to Shapiro et al., Bakker et al. found that initial blood
lactate levels were not predictive of outcome, but the time to decrease lactate levels
was more indicative of outcome. The role of lactate in predicting mortality is still
not fully understood.
2.4
Thesis Goals
First, we were interested in quantifying the relationship between lactate and other
easily measured parameters in ICU patients to access the predictive value of initial
lactate levels in the ICU. More concretely, we wanted to more fully understand the
association between initial lactate levels and ICU mortality, 30 day mortality, length
of stay, and severity of illness. We wanted to quantify differences in patient outcome
associated with differing levels of initial lactate. We focused on three different ranges
of lactate levels: low (below 2.0 mmol/L), medium (between 2.0 mmol/L and 4.0
mmol/L), and high (above 4.0 mmol/L).
Our next goal was to identify sub-populations whose lactate levels were more
associated with mortality and severity of illness. In order to achieve this goal, we
found which conditions were most commonly associated with high lactate levels. We
hypothesized that patients in certain care units might be more likely to have high
lactate levels than others. To test this hypothesis, we specifically analyzed lactate
associations with ICU mortality, 30 day mortality, length of stay, and severity of illness
in each care unit of the ICU separately. Lastly, we analyzed the role of lactate in
sepsis patients. Because there are multiple ways to define sepsis patients, we identified
three different cohorts of patients with sepsis and analyzed their lactate levels.
Finally we wanted to determine if knowing change in lactate gave additional predictive value to the initial lactate value. We focused on the implications of lactate
change in combination with the level of the initial lactate on ICU mortality. Previous
studies showed that change in lactate is important for predicting patient outcome, so
24
we were interested to see the effect of controlling for the initial lactate measurement.
25
26
Chapter 3
Methods
We studied the association of lactate levels with ICU mortality, 30 day mortality,
length of stay, and illness severity. In addition, we studied how lactate clearance
rates were associated with ICU mortality. We analyzed the associations of lactate in
all ICU patients and also in the sub-cohort of septic patients because of the previous
studies investigating the association of lactate levels to patient outcome. The goal
of this study was to better understand the predictive value of lactate in an intensive
care setting.
3.1
Cohort
Patient information used in this investigation was provided by the Multiparameter
Intelligent Monitoring in Intensive Care (MIMIC) II database (version 2.6) [9]. The
online database contains records of 32,536 patients admitted to the ICU of Beth Israel
Deaconess Medical Center in Boston, Massachusetts. There is detailed information
about each of the 40,426 ICU stays including care unit, lab work, vitals, nursing notes,
etc. which can be extracted for research and analysis.
27
3.2
Initial Lactate Measurements
Initially we identified which patients (based on their ICU stay ID) had lactate measurements made during their ICU admission. Using the Lab Events table in the
MIMIC II database, we found the first lactate measurement for each ICU admission.
We also identified patients who did not have lactate measurements during their ICU
stays. We looked specifically at the initial lactate measurement to understand further
if it had any predictive value. We found 14,402 ICU admissions in which the patient
had lactate measured, and 26,024 admissions where no measurement was taken.
Based on previous studies, we divided the group of ICU admissions into three
groups based on their initial lactate measurements: lactate less than 2.0 mmol/L,
between 2.0 mmol/L and 4.0 mmol/L, and greater than 4.0 mmol/L. Lactate levels
under 2.0 mmol/L are considered normal whereas lactate levels of 4.0 mmol/L or
greater are considered very high. There were 8,107 admissions with initial lactate
levels less than 2.0 mmol/L, 4,055 admissions with lactate levels between 2.0 mmol/L
and 4.0 mmol/L, and 2,240 admissions with lactate levels above 4.0 mmol/L. The
cohort breakdown is shown in Table 2.1.
Table 3.1: Cohort broken down into groups by lactate measurement
Initial Lactate Value
Number of ICU Admissions
Percent of Cohort
≤2.0 mmol/L
8,107
20.1%
> 2mmol/L and < 4.0 mmol/L
4,055
10.1%
≥ 4.0 mmol/L
2,240
5.5%
26,024
64.3 %
No Lactate Measurement Taken
28
3.3
3.3.1
Parameters Compared to Lactate
ICU Mortality
To determine ICU mortality, we used the ICU expire flag in the ICU Stay Events
Detail table. We compared the lactate measurement and ICU expire flag based on
a shared ICU stay ID. The ICU mortality rate was the number of ICU admissions
where the patient died in the ICU compared to the overall number of ICU admissions.
3.3.2
30 Day Mortality
For 30-day mortality, first we checked if the patient had a date of death on record. If
the patient’s date of death was within 30 days of their discharge from the ICU, they
were counted as dead within 30 days.
3.3.3
Length of Stay
When determining length of stay, we found the average and the median length of
stay of all of the ICU admissions corresponding to the ICU stay IDs in each group of
patients we considered. We only included patients who did not die during their ICU
stay. We looked only at survivors in order to use length of stay as a surrogate for
how sick a patient was.
3.3.4
Severity of Illness
In order to understand the association between lactate levels and severity of illness,
we looked at the first SAPS I and first SOFA severity scores of each ICU admission
and compared them to the initial lactate level. The SAPS I score is used to classify
how likely a patient is to die while in the ICU [10]. The SAPS I score is based on 14
measured clinical variables and is not based on diagnosis. The SOFA score is another
prognostic indicator used in the ICU [11]. The SOFA score was originally created in
order to assess the organ failure status of a patient.
29
3.4
Conditions Associated with High Lactate
Patients who enter the hospital are billed according to the illnesses they are treated
for and the procedures they have done while at the hospital. The system in place
in the United States to assign codes to diseases and procedures is the International
Classification of Diseases, Ninth Revision (ICD-9) [12]. These codes are used to
determine how a patient will be billed. In retrospective research studies, these billing
codes are often used to see why a patient was in the hospital and what the major
conditions were for which they were treated while at the hospital. We identified the
ten most common primary ICD-9 codes associated with ICU admissions that had an
initial lactate greater than 4.0 mmol/L. We wanted to gain a better understanding
of the main causes for high lactate levels.
3.5
Change in Lactate
In addition to looking at initial lactate measurements,we also compared change in
lactate level to ICU mortality in order to gain a better understanding of why some
people with high initial lactate died while others did not. In previous studies, evidence showed that high initial lactate levels were associated with higher rates of
ICU mortality so we decided to look at the relationship between high initial lactate
and change in lactate. We determined the percent change in lactate from the initial
lactate measurement to the second measurement for patients who had their lactate
measured two or more times throughout their ICU stay. We found the first two lactate
measurements for each ICU stay and calculated clearance as:
(𝐿𝑎𝑐𝑡𝑎𝑡𝑒𝑠𝑒𝑐𝑜𝑛𝑑 − 𝐿𝑎𝑐𝑡𝑎𝑡𝑒𝑓 𝑖𝑟𝑠𝑡 )/(𝐿𝑎𝑐𝑡𝑎𝑡𝑒𝑓 𝑖𝑟𝑠𝑡 ) * 100
We compared ICU mortality rate to the change in lactate to try to better understand
why some people with high initial lactate levels die while others live.
30
3.6
Septic Cohort
To further investigate lactate’s predictive value, we looked specifically at septic patients because of previous associations made between sepsis and lactate. In order to
characterize the septic population, we identified three different septic cohorts by using
different criteria. We used both Angus and Martin criteria for identifying patients
with severe sepsis. In addition we identified a septic cohort based on the presence of
infection and systemic inflammatory response syndrome.
3.6.1
Martin Implementation
The Martin definition of severe sepsis is based on ICD-9 codes for sepsis as well
as ICD-9 codes and procedure codes for organ dysfunction [13]. To be considered
severely septic, the patient needs an ICD-9 code for septicemia (038), septicemic
(020.0), bacteremia (790.7), disseminated fungal infection (117.9), disseminated candida infection (112.5), or disseminated fungal endocarditis (112.81). In addition to
one of the previous codes, the patient needs a code representing organ dysfunction.
Because a patient may only be billed for a set number of conditions during each
hospital visit, some septic patients may not have a billing code specifically for sepsis.
If the patient has other serious conditions, the assigned codes may not include sepsis
even when the patient had the diagnosis during their hospital stay.
3.6.2
Angus Implementation
The Angus implementation for identifying severely septic patients also uses ICD-9
codes to determine how a patient was diagnosed while at the hospital. Instead of
using codes specifically for sepsis, the implementation finds patients who have a code
for a bacterial or fungal infection and acute organ dysfunction – two of the main
criteria for diagnosing severe sepsis [5]. To be identified as severely septic in our
cohort, a patient needed to have at least one ICD-9 code from each category of codes
(infection and organ dysfunction). The Angus implementation of the definition of
severe sepsis has been found to be reasonably accurate at identifying patients with
31
severe sepsis, but it is imperfect[14].
3.6.3
Infection and SIRS
We defined a cohort based on the definition of sepsis, rather than based on ICD-9
codes. We found patients that had a confirmed or suspected infection as well as two
out of the four SIRS criteria, defined in Chapter One. We decided to look at just
patients who were septic at the onset of their ICU admission as an ad hoc decision
because it was simplest.
To determine either confirmed or suspected infection, we looked at whether the
patient was started on IV antibiotics within the first 24 hours of admission to the
ICU. The list of antibiotics we considered are located in section A.1.
We also determined whether the patient had two of the four SIRS criteria at
relatively the same time during the first day of the ICU stay. We found each time the
patient had their heart rate, temperature, white blood cell count, or respiratory rate
measured. If two of the patient’s measurements met the criteria and were measured
within eight hours of each other, the patient was considered to have SIRS.
If the patient had SIRS and had a confirmed or suspected infection on the first
day, we considered the patient to have sepsis on the first day of admission. When
comparing lactate measurements within this cohort, we only looked at initial lactate
measurements taken on the first day of the ICU stay, to remain consistent with how
the cohort was chosen. Therefore, the lactate measurements were taken at the same
time as the suspected sepsis.
We evaluated our selected cohort by examining ten randomly selected cases from
the cohort. We looked at the nursing notes and discharge summaries associated with
each of the ten admissions, and we found in each case evidence of sepsis or of the
implementation of the septic protocol.
The criteria used to find this cohort is the least stringent of the three methods
because it may capture patients who had suspected sepsis but in the end did not
actually have sepsis.
32
Chapter 4
Lactate Associations in All ICU
Patients
4.1
Associations with Initial Lactate Levels
There were 40,426 ICU admissions in total in our cohort. Of these admissions, there
were 14,402 ICU admissions that had lactate measurements (35.6%) and 26,024 ICU
admissions that did not have lactate measurements (64.4%).
4.1.1
ICU Mortality Rate
Compared to the overall ICU mortality of all ICU admissions, 5.8%, the ICU mortality
rate of ICU admissions that had lactate measured was much higher at 13.1%. Those
who did not receive a lactate measurement had a very low ICU mortality rate of 1.8%.
This may suggest that only patients who were very ill had their lactate measured.
All ICU mortality rate results are detailed in Table 4.1.
We found that as initial lactate levels increased, so did the ICU mortality rate.
Notice in Table 4.1 that admissions with normal lactates (lactate level less than or
equal to 2.0 mmol/L), the ICU mortality rate was only 9.6% whereas admissions with
high lactate (lactate level greater than 4.0 mmol/L) had a much larger ICU mortality
rate of 26.8%. The ICU mortality rate of those with initial lactate levels between
33
2.0 mmol/L and 4.0 mmol/L was 12.5%, which was higher than that of those with
normal lactate levels, but much lower than that of patients with very high lactate
levels.
Table 4.1: ICU Mortality Vs. Lactate for All ICU Admissions
Died in
the ICU
Did not
Die in the
ICU
Total ICU
admissions
ICU
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
776
7,331
8,107
9.60%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
506
3,549
4,055
12.50%
Initial Lactate
≥ 4.0 mmol/L
600
1,640
2,240
26.80%
Any Lactate
Measurement
1,882
12,520
14,402
13.10%
No Lactate
Measurement
469
25,555
26,024
1.80%
2,351
38,075
40,426
5.80%
All ICU
Admissions
4.1.2
30 Day Mortality Rate
In general, the 30 day mortality rates were approximately twice the ICU mortality
rates, but they still followed the same trend. The 30 day mortality rate of those
with any lactate measurement was much higher than the rate of all ICU admissions
whereas the 30 day mortality rate of those without a lactate measurement was much
lower. Admissions with a lactate measurement had a 30 day mortality rate of 22.5%,
compared to those without a lactate measurement which had a 30 day mortality rate
of only 5.9%. Again, the results suggest that patients who were not as sick did not
34
get their lactate measured. All 30 day mortality results are detailed in Table 4.2.
As initial lactate level increased, so did the 30 day mortality rate. Admissions with
initial lactate less than 2.0 mmol/L had a 30 day mortality rate of 17.5%. Admissions
with initial lactate above 4.0 mmol/L had the highest 30 day mortality rate of 41.6%.
Table 4.2: 30 Day Mortality Vs. Lactate for All ICU Admissions
Died
within 30
Days
Did not
Die within
30 Days
Total ICU
admissions
30 Day
Mortality
Rate
1,416
6,691
8,107
17.5%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
885
3,170
4,055
21.8%
Initial Lactate
≥ 4.0 mmol/L
932
1,308
2,240
41.6%
Any Lactate
Measurement
3,233
11,169
14,402
22.5%
No Lactate
Measurement
1,817
24,207
26,024
7.0%
All ICU
Admissions
5,048
35,378
40,426
12.5%
Initial Lactate
≤ 2.0 mmol/L
4.1.3
Length of Stay
Compared to length of stay of all ICU admissions who survived which had an average
of 5.4 days and a median of 2.0 days, admissions with any lactate measurement had
a longer length of stay than those without a lactate measurement. This may suggest
that admissions with lactate measurements were sicker on average than those who
did not. All length of stay results of ICU survivors are summarized in Table 4.3.
ICU admissions in which the patient survived their ICU stay, who had an initial
35
lactate level below 4.0 mmol/L, all had similar lengths of stay. The average and
median length of stay for those with initial lactate below 2.0 mmol/L was 6.6 and 3.4
days respectively, and the average and median length of stay for those with initial
lactate between 2.0 mmol/L and 4.0 mmol/L was 6.6 and 3.2 days. Admissions with
initial lactate above 4.0 mmol/L had a slightly longer length of stay with an average
of 7.4 days and a median of 3.9 days.
Table 4.3: Average and Median Length of Stay Vs. Lactate for All ICU Admissions
Who Survived Their ICU Stay
Average Length of
Stay (Days)
Median Length of
Stay (Days)
Initial Lactate
≤ 2.0 mmol/L
6.6
3.4
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
6.6
3.2
Initial Lactate
≥ 4.0 mmol/L
7.4
3.9
Any Lactate
Measurement
6.9
3.4
No Lactate
Measurement
4.7
1.5
All ICU Admissions
5.4
2.0
4.1.4
Severity of Illness
As a surrogate for severity of illness, we determined the average SOFA score and
SAPS I scores of ICU admissions with lactate levels. Admissions with any lactate
measurement had much higher severity scores than admissions without a lactate measurement, as seen in Table 4.4. We saw greater likelihood of mortality, based on these
scores, in patients with higher lactate levels. Admissions with initial lactate levels less
36
than 2.0 mmol/L had an average SOFA score of 6.50 and an average SAPS I score of
14.88. Admissions with high initial lactate levels above 4.0 mmol/L had the highest
average severity scores, a 10.05 SOFA score and a 19.23 SAPS I score. The average
SOFA score and SAPS I score for all ICU admissions with a lactate measurement
were 7.41 and 15.91, respectively. Table 4.4 summarizes the average severity scores.
Table 4.4: Average SOFA and SAPS I Scores Vs. Lactate for All ICU Admissions
Average SOFA Score
Average SAPS I Score
Initial Lactate
≤ 2.0 mmol/L
6.50
14.88
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
7.75
16.18
Initial Lactate
≥ 4.0 mmol/L
10.05
19.23
Any Lactate
Measurement
7.41
15.91
No Lactate
Measurement
3.91
11.6
All ICU Admissions
5.54
13.7
4.1.5
Change in Lactate
When initial lactate was below 2.0 mmol/L, we saw a slight increase in ICU mortality
when the second lactate level increased, but the ICU mortality rate stayed under 20%
for all changes in lactate level as shown in Figure 4-1. When initial lactate levels were
between 2.0 mmol/L and 4.0 mmol/L, there was a greater increase in ICU mortality
rate when lactate increased rather than decreased as shown in Figure 4-2. An increase
in lactate of more than 80% had an ICU mortality rate of 34.62%.
The greatest difference in ICU mortality rates was seen in admissions with initial
37
lactate greater than 4.0 mmol/L. The ICU mortality rate continuously increased as
the percent change in lactate increased as seen in Figure 4-3. When lactate levels
decreased by more than 80%, the ICU mortality rate was only 11.76%, but when the
lactate levels increased by more than 80%, the ICU mortality rate was 65.38%. The
extent by which the lactate level changed from the first measurement to the second
measurement when the initial measurement was above 4.0 mmol/L greatly influenced
whether a patient died. Figure 4-4 shows the trend of ICU mortality rates compared
to percent change in lactate of admissions with any initial lactate levels.
Figure 4-1: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Below 2.0 mmol/L
38
Figure 4-2: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Between 2.0 mmol/L and 4.0 mmol/L
Figure 4-3: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L
39
Figure 4-4: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality for All ICU Admissions with Any Lactate
4.2
Lactate Associations in Different Care Units
To further understand lactate associations in ICU patients, we looked at the ICU
mortality, 30 day mortality, and length of stay in each individual care unit. There
were 3980 admissions in the MICU with lactate measurements, 1771 in the FICU,
2590 in the CCU, 5149 in the CSRU, and 869 in the SICU. We were interested to
determine if patients in a particular care unit had greater associations with lactate.
4.2.1
ICU Mortality
We looked at the ICU mortality rate of admissions with lactate measurements in
each of the different care units. The ICU mortality rate of admissions with lactate
measurements for each care unit are detailed in Table 4.5. The ICU mortality rates
of each care unit were higher when only admissions with any lactate measurement
were included. The ICU mortality rate including those with or without lactate measurements was 9.5% in the MICU, 8.1% in the FICU, 7.5% in the CCU, 4.7% in the
40
CSRU, and 6.0% in the SICU.
The ICU mortality rate was the lowest in the CSRU, most likely because cardiac
surgery recovery patients are generally healthy before going into surgery. The cause
for increased lactate during anesthesia, surgery, and by-pass is quickly resolvable
as opposed to the causes of critical medical illnesses. The ICU mortality rate may
also be lower for patients in the CSRU because there is a protocol in place that
requires lactate measurements to be taken in the CSRU (at the hospital from which
we get our patient data). The ICU mortality rate was the highest in the MICU most
likely because patients in the MICU have underlying illnesses. Admissions with a
lactate measurement on average must be sicker than those who do not get lactate
measurements.
Table 4.5: ICU Mortality of Admissions With a Lactate Measurement in Each ICU
Care Unit
Died in
the ICU
Did not
Die in the
ICU
Total
Admissions
ICU
Mortality
Rate
MICU
704
3,276
3,980
17.7%
FICU
231
1,540
1,771
13.0%
CCU
420
2,170
2,590
16.2%
CSRU
438
4,711
5,149
8.5%
SICU
85
784
869
9.8%
4.2.2
30 Day Mortality
We looked at the 30 day mortality rate of admissions with lactate measurements in
each care unit. The 30 day mortality rates of those with lactate measurements in
each care unit are detailed in Table 4.6. The 30 day mortality rates of just patients
with lactate measurements were much higher than the 30 day mortality rates of each
41
care unit when all patients were included. The 30 day mortality rate including those
with or without lactate measurements was 21.6% in the MICU, 15.0% in the FICU,
15.3% in the CCU, 11.0% in the CSRU, and 11.9% in the SICU.
Table 4.6: 30 Day Mortality of Admissions With a Lactate Measurement in Each ICU
Care Unit
Died
within 30
Days
Did not
Die within
30 Days
Total
Admissions
30 Day
Mortality
Rate
MICU
1,249
2,731
3,980
31.4%
FICU
360
1,411
1,771
20.3%
CCU
657
1,933
2,590
25.4%
CSRU
819
4,330
5,149
15.9%
SICU
142
727
869
16.3%
4.2.3
Length of Stay
The average and median length of stay for admissions with lactate measurements
that did not die in the ICU was similar across all care units. The average and median
lengths of stay for each care unit are summarized in Table 4.7.
4.3
Conditions Associated with High Lactate
We identified which conditions were most associated with high lactate levels. We
found the primary ICD-9 codes of all ICU admissions with an initial lactate level
greater than 4.0 mmol/L. The ten most frequent conditions were unspecified septicemia (38.9), coronary atherosclerosis of native coronary artery (414.01), subendocardial infarction initial episode of care (410.71), septicemia due to E. coli (38.42),
42
Table 4.7: Average and Median Length of Stay of Admissions With Lactate Levels
Broken Down By Care Unit
Average Length of
Stay (Days)
Median Length of
Stay (Days)
MICU
6.5
3.4
FICU
7.5
3.7
CCU
7.3
4.0
CSRU
6.1
3.1
SICU
5.5
2.8
acute respiratory failure (518.81), congestive heart failure unspecified (428), pneumonitis due to inhalation of food or vomitus (507), acute and subacute necrosis of
liver (570), acute myocardial infarction of other anterior wall (410.11), and acute
myocardial infarction of other inferior wall (410.41). The number of admissions with
initial lactate levels above 4.0 mmol/L with one of the primary ICD-9 codes listed is
shown in Table 4.8.
We found that the most common conditions associated with lactate can be broken in three categories: sepsis, cardiac conditions, and respiratory conditions. The
condition most commonly associated with high lactate was sepsis, so we have further
studied this population in the following chapter. Further research should be done on
the other two populations of patients associated with high lactate levels. Even though
these populations were the most common that we found, there were many different
conditions that must be associated with high lactate. There were 2,240 admissions
with high lactate levels, but the condition most associated with high lactate was only
found as the primary condition in 333 admissions.
43
Table 4.8: Primary ICD-9 Codes Most Commonly Associated with High Lactate
Levels (Of a Total of 2,240 Admissions With High Lactate Levels)
Description
Number of
Admissions
unspecified septicemia, septicemia due to
E. coli
333
acute myocardial infarction of other
anterior/inferior wall, subendocardial
infarction initial episode of care
129
414.01
coronary atherosclerosis of native
coronary artery
89
518.81
acute respiratory failure
41
428
congestive heart failure unspecified
34
507
pneumonitis due to inhalation of food or
vomitus
28
570
acute and subacute necrosis of liver
26
ICD-9 Code
38.9, 38.42
410.11, 410.41,
410.71
44
Chapter 5
Lactate Associations in Septic
Patients
5.1
Martin Criteria Severe Sepsis
There were 3,958 ICU admissions which met the Martin criteria for severe sepsis.
No lactate measurement was taken during 938 of these admissions (23.7%), whereas
lactate measurements were taken during 3,021 of these admissions (76.3%).
5.1.1
ICU Mortality
Martin criteria severe sepsis admissions with no lactate measurement had a much
lower ICU mortality rate of 8.5% compared to the 35.3% ICU mortality rate of admissions with a lactate measurement. Due to the 938 patients without any lactate
measurements, the overall ICU mortality rate of all severely septic patients under
Martin criteria was 21.5%. Table 5.1 summarizes the results in detail.
Notice in Table 5.1 that ICU mortality increased as the initial lactate level increased. Martin criteria severe sepsis admissions with lactate levels below 2.0 mmol/L
had an 18.9% ICU mortality rate, compared to patients with initial lactate levels
above 4.0 mmol/L which had a much larger ICU mortality rate of 46.6%.
45
Table 5.1: ICU Mortality Vs. Lactate for Martin Criteria Severe Sepsis Patients
Died in
the ICU
Did not
Die in the
ICU
Total ICU
admissions
ICU
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
296
1,272
1,568
18.9%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
221
646
867
25.5%
Initial Lactate
≥ 4.0 mmol/L
272
314
586
46.6%
Any Lactate
Measurement
789
2,232
3021
35.3%
No Lactate
Measurement
60
878
938
8.5%
All Sepsis
Admissions
849
3,109
3,958
21.5%
5.1.2
30 Day Mortality
The 30 day morality rate of all severely septic patients identified with the Martin
criteria was 36.2%. Admissions where no lactate was measured had a slightly lower
30 day mortality rate of 25.8%, and those where lactate was measured had a slightly
higher 30 day mortality rate of 39.5%. Table 5.2 summarizes the 30 day mortality
results.
Just as ICU mortality rate increased with increased lactate levels, so did the 30 day
mortality rate. Severely septic patients with initial lactate levels below 2.0 mmol/L
had a 30 day mortality rate of 31.6%, whereas admissions with lactate levels above
4.0 mmol/L had a very high 30 day mortality rate of 59.2%.
46
Table 5.2: 30-Day Mortality Vs. Lactate for Martin Criteria Severe Sepsis Patients
Died
within 30
Days
Did not
Die within
30 Days
Total ICU
admissions
30 Day
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
496
1,072
1,568
31.6%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
350
517
867
40.4%
Initial Lactate
≥ 4.0 mmol/L
346
239
585
59.2%
Any Lactate
Measurement
1,192
1,828
3,020
39.5%
No Lactate
Measurement
242
696
938
25.8%
1,434
2,524
3,958
36.2%
All Sepsis
Admissions
5.1.3
Length of Stay
The average and median length of stay of ICU survivors, 8.8 and 4.1 days respectively,
was much longer than the average and median length of stay of all ICU admissions
who survived, 5.4 and 2.0 days respectively. Those without lactate measurements
had average and median lengths of stay of 5.0 and 2.1 days respectively, which was
almost identical to the average and median length of stay of all ICU admissions who
survived. Admissions that had a lactate measured had very high average and median
lengths of stay that were 10.2 and 5.6 days, respectively. The length of stay results
for Martin criteria sepsis patients who survived their ICU stay are summarized in
Table 5.3.
The average and median length of stay of Martin criteria severe sepsis patients
who survived their ICU stay increased as the initial lactate measurement increased.
47
Those with low lactate levels had an average length of stay of 9.7 days and a median
length of stay of 5.2 days, whereas those with very high lactate levels stayed in the
ICU for an average of 11.5 days and a median of 7.6 days.
Table 5.3: Average and Median Length of Stay for Sepsis Admissions Defined by
Martin Criteria Who Survived Their ICU Stay
Average Length of
Stay (Days)
Median Length of
Stay (Days)
Initial Lactate
≤ 2.0 mmol/L
9.7
5.2
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
10.6
5.6
Initial Lactate
≥ 4.0 mmol/L
11.5
7.6
Any Lactate
Measurement
10.2
5.6
No Lactate
Measurement
5.0
2.1
All Sepsis Admissions
8.8
4.1
5.1.4
Severity of Illness
The average SOFA score and the average SAPS I score both increased as the initial
level of lactate increased. Admissions with any lactate measurement had much higher
scores on average than admissions without a lactate measurement. Martin criteria
severe sepsis admissions with any lactate level had an average SOFA score of 9.04 and
the average SAPS I score of 17.33. Admissions with lactate below 2.0 mmol/L had
average SOFA score of 7.69 and SAPS I score of 15.94. There was a large increase
in average severity scores in admissions with very high lactate levels greater than 4.0
48
mmol/L. The average SOFA score was 12.02 and the average SAPS I score was 20.75.
The severity score results are summarized in Table 5.4.
Table 5.4: Average SOFA and SAPS I Scores Sepsis Admissions Defined by Martin
Criteria
Average SOFA Score
Average SAPS I Score
Initial Lactate
≤ 2.0 mmol/L
7.69
15.94
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
9.49
17.57
Initial Lactate
≥ 4.0 mmol/L
12.02
20.75
Any Lactate
Measurement
9.04
17.33
No Lactate
Measurement
5.31
12.96
All Sepsis Admissions
8.20
16.44
5.1.5
Change in Lactate
The ICU mortality rate of Martin criteria severe sepsis admissions with initial lactate
greater than 4.0 mmol/L was much lower when the lactate decreased from the first
to the second measurement than when the lactate increased, as shown in Figure 5-1.
When lactate decreased by more than 80%, the mortality rate was 28.57%. When
lactate increased by 40 to 60%, the mortality rate was very high at 76.47%. The
mortality rate was lower when lactate increased by more than 80%, but this was
mostly likely due to the small sample size of admissions with initial lactate levels
above 4.0 mmol/L and increasing by more than 80%. Overall, the mortality rate was
much larger when lactate increased.
49
Figure 5-1: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L for Sepsis Admissions Defined by Martin Criteria
5.2
Angus Criteria Severe Sepsis
We identified 6,362 admissions that met the Angus criteria for severe sepsis. Of these
admissions, 4,320 had lactate levels (67.9%) and 2,043 did not have lactate levels
(32.1%). We found more patients with severe sepsis using the Angus criteria than we
found using the Martin criteria because the Angus criteria are less stringent.
5.2.1
ICU Mortality
The ICU mortality rate of all admissions identified using the Angus criteria for severe
sepsis was 16.9%. Admissions with a lactate level had a mortality rate of 22.5%,
whereas admissions without a lactate level had a much lower mortality rate of 5.1%.
Those without lactate levels seem to be less ill on average than those with lactate
levels.
The mortality rate increased as the initial lactate level increased. The ICU mortality rate of admissions with initial lactate less than 2.0 mmol/L was 16.8%, whereas
50
the ICU mortality of admissions with initial lactate greater than 4.0 mmol/L was
39.5%. These results are detailed in Table 5.5.
Table 5.5: ICU Mortality Vs. Lactate for Sepsis Admissions Defined by Angus Criteria
Died in
the ICU
Did not
Die in the
ICU
Total ICU
admissions
ICU
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
396
1,967
2,363
16.8%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
263
906
1169
22.5%
Initial Lactate
≥ 4.0 mmol/L
312
476
788
39.5%
Any Lactate
Measurement
971
3,349
4,320
22.5%
No Lactate
Measurement
104
1939
2,043
5.1%
1,075
5,288
6,362
16.9%
All Sepsis
Admissions
5.2.2
30 Day Mortality
The 30 day mortality rate of all admissions identified by the Angus criteria for severe
sepsis was 30.0%. Admissions with a lactate measurement had a slightly higher 30
day mortality rate of 34.5%, whereas those without a lactate measurement had a
lower 30 day mortality rate of 20.4%.
The 30 day mortality rate of admissions with high lactate levels above 4.0 mmol/L,
53.3%, was much higher than the 30 day mortality rate of admissions with low lactate
below 2.0 mmol/L, 27.6%. Details for 30 day mortality are located in Table 5.6.
51
Table 5.6: 30-Day Mortality Vs. Lactate for Sepsis Admissions Defined by Angus
Criteria
Died
within 30
Days
Did not
Die within
30 Days
Total ICU
admissions
30 Day
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
653
1,710
2,363
27.6%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
419
750
1169
35.8%
Initial Lactate
≥ 4.0 mmol/L
420
368
788
53.3%
Any Lactate
Measurement
1,492
2,828
4,320
34.5%
No Lactate
Measurement
416
1,627
2,043
20.4%
1,908
4,455
6,363
30.0%
All Sepsis
Admissions
5.2.3
Length of Stay
The average and median length of stay of all Angus criteria sepsis admissions that
did not die while in the ICU was 7.1 days and 3.5 days, respectively. The admissions
that had a lactate level had an average length of stay of 9.3 days and a median length
of stay of 5.1 days. Admissions without a lactate level had a much shorter length of
stay on average. Admissions without lactate levels had an average length of stay of
3.1 days and a median length of stay of 2.0 days.
The length of stay was longer for admissions with higher initial lactate levels.
Admissions with lactate levels below 2.0 mmol/L had an average length of stay of 8.9
days and a median of 4.8 days. Admissions with high lactate above 4.0 mmol/L had
a longer average length of stay of 10.9 days and a median of 6.9 days. The results are
52
summarized in Table 5.7.
Table 5.7: Average and Median Length of Stay Vs. Lactate for Sepsis Admissions
Defined by the Angus Criteria Who Survived Their ICU Stay
Average Length of
Stay (Days)
Median Length of
Stay (Days)
Initial Lactate
≤ 2.0 mmol/L
8.9
4.8
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
9.5
5.0
Initial Lactate
≥ 4.0 mmol/L
10.9
6.9
Any Lactate
Measurement
9.3
5.1
No Lactate
Measurement
3.1
2.0
All Sepsis Admissions
7.1
3.5
5.2.4
Severity of Illness
The average SOFA score and average SAPS I score of all severely septic patients identified using the Angus criteria that had lactate levels were 8.60 and 16.96, respectively.
Admissions with any lactate measurement had higher severity scores on average than
admissions without a lactate measurement. The severity of illness increased as the
initial lactate level increased. Admissions with initial lactate less than 2.0 mmol/L
had an average SOFA score of 7.40 and an average SAPS I score of 15.70. Admissions
with high lactate above than 4.0 mmol/L had the highest average SOFA score of
11.62 and the highest average SAPS I score of 20.40. These results are summarized
in Table 5.8.
53
Table 5.8: Average SOFA and SAPS I Scores Vs. Lactate for Sepsis Admissions
Defined by Angus Criteria
Average SOFA Score
Average SAPS I Score
Initial Lactate
≤ 2.0 mmol/L
7.40
15.70
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
8.98
17.21
Initial Lactate
≥ 4.0 mmol/L
11.62
20.40
Any Lactate
Measurement
8.60
16.96
No Lactate
Measurement
4.73
12.54
All Sepsis Admissions
7.38
15.66
5.2.5
Change in Lactate
The ICU mortality rate of Angus criteria severe sepsis admissions with initial lactate
greater than 4.0 mmol/L was much lower when the lactate decreased from the first
to the second measurement than when the lactate increased, as shown in figure 52. Although the ICU mortality rate is not strictly increasing with increased percent
change in lactate, the overall trend is increasing. When lactate decreased by more
than 80%, the ICU mortality rate was very low at 9.09%. When lactate increased by
60 to 80%, the ICU mortality rate was the highest at 77.78%. Overall, the mortality
rate was much larger when lactate levels increased.
54
Figure 5-2: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L for Sepsis Patients
Defined by Angus Criteria
5.3
Infection and SIRS Criteria Sepsis
There were 9,911 ICU admissions identified as septic on the first day of admission,
using the criteria of having an infection and SIRS, as described in section 3.6.3. Of
these admissions, 6,032 had lactate measurements on the first day of their admission
(60.9%) whereas 3,879 admissions did not (39.1%).
5.3.1
ICU Mortality
Admissions that were identified as septic on the first day using the infection and
SIRS criteria had an overall ICU mortality rate of 11.7%. The ICU mortality rate
of admissions with lactate levels was higher (15.1%), and the ICU mortality rate of
those without lactate levels was much lower (6.3%).
The ICU mortality rate of admissions with initial lactate levels measured on the
first day below 4.0 mmol/L was low. Admissions initial lactate below 2.0 mmol/L
had an ICU mortality rate of 10.9%, and admissions with initial lactate level between
55
2.0 mmol/L and 4.0 mmol/L had an ICU mortality rate of 12.9%. Admissions with
initial lactate greater than 4.0 mmol/L had a much higher ICU mortality rate of
30.2%. These results are detailed in Table 5.9.
Table 5.9: ICU Mortality Vs. Lactate for Sepsis Patients Defined by Infection and
SIRS Criteria
Died in
the ICU
Did not
Die in the
ICU
Total ICU
admissions
ICU
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
333
2,732
3,065
10.9%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
236
1,601
1,837
12.9%
Initial Lactate
≥ 4.0 mmol/L
341
789
1,130
30.2%
Any Lactate
Measurement
910
5,122
6,032
15.1%
No Lactate
Measurement
246
3,633
3,879
6.3%
1,156
8,755
9,911
11.7%
All Sepsis
Admissions
5.3.2
30 Day Mortality
The overall 30 day mortality rate of sepsis admissions found using the infection and
SIRS criteria was 21.2%. Admissions without a lactate measurement on the first
day of admission had a 30 day mortality rate of 15.3%, whereas those with a lactate
measurement on the first day had a 30 day mortality rate of 25.1%.
The 30 day mortality rate for admissions with lactate below 4.0 mmol/L was much
smaller than the 30 day mortality rate for admissions with lactate above 4.0 mmol/L.
56
Admissions with initial lactate less than 2.0 mmol/L had a 30 day mortality rate of
19.8%, and admissions with initial lactate between 2.0 mmol/L and 4.0 mmol/L had
a 30 day mortality rate of 22.8%. Admissions with lactate above 4.0 mmol/L had a
high 30 day mortality rate of 42.9%. Details of these results are found in Table 5.10.
Table 5.10: 30 Day Mortality Vs. Lactate for Sepsis Patients Defined by Infection
and SIRS
Died
within 30
Days
Did not
Die within
30 Days
Total ICU
admissions
30 Day
Mortality
Rate
Initial Lactate
≤ 2.0 mmol/L
607
2,458
3,065
19.8%
Initial Lactate
> 2.0 mmol/L
and < 4.0 mmol/L
419
1,418
1,837
22.8%
Initial Lactate
≥ 4.0 mmol/L
485
645
1,130
42.9%
Any Lactate
Measurement
1,511
4,521
6,032
25.1%
No Lactate
Measurement
594
3,285
3,879
15.3%
2,105
7,806
9,911
21.2%
All Sepsis
Admissions
5.3.3
Length of Stay
The average and median length of stay of all sepsis admissions defined by infection
and SIRS on the first day that did not die while in the ICU was 5.4 days and 2.9 days,
respectively. The admissions with a lactate level had an average length of stay was
6.5 days and a median length of stay was 3.4 days. The admissions without a lactate
level had a much shorter length of stay on average. Those without lactate levels had
57
an average length of stay was 3.9 days and a median length of stay of 2.3 days.
The length of stay was slightly longer for admissions with higher initial lactate
levels. Admissions with lactate levels below 2.0 mmol/L had an average length of stay
of 6.0 days and a median of 3.2 days, whereas admissions with high lactate above 4.0
mmol/L had a longer average length of stay of 7.7 days and a median of 4.5 days.
The results are summarized in Table 5.11.
Table 5.11: Average and Median Length of Stay for Sepsis Admissions Defined by
Infection and SIRS Who Survived Their ICU Stay
Average Length of
Stay (Days)
Median Length of
Stay (Days)
Initial Lactate
≤ 2.0 mmol/L
6.0
3.2
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
6.7
3.3
Initial Lactate
≥ 4.0 mmol/L
7.7
4.5
Any Lactate
Measurement
6.5
3.4
No Lactate
Measurement
3.9
2.3
All Sepsis Admissions
5.4
2.9
5.3.4
Severity of Illness
The average SOFA score and average SAPS I score of all septic patients defined by
infection and SIRS with lactate levels were 8.42 and 16.8, respectively. Admissions
with any lactate measurement on the first day of admission had higher severity scores
on average than admissions that did not have a lactate measurement on the first day.
58
The severity of illness increased as the initial lactate level increased. Admissions with
initial lactate less than 2.0 mmol/L had an average SOFA score of 7.4 and an average
SAPS I score of 15.79. Admissions with high lactate above than 4.0 mmol/L had the
highest average SOFA score of 11.18 and the highest average SAPS I score of 19.88.
These results are summarized in Table 5.12.
Table 5.12: Average SOFA and SAPS I Scores Vs. Lactate for Sepsis Admissions
Defined by Infection and SIRS Criteria
Average SOFA Score
Average SAPS I Score
Initial Lactate
≤ 2.0 mmol/L
7.4
15.79
Initial Lactate
>2.0 mmol/L and
<4.0 mmol/L
8.64
16.83
Initial Lactate
≥ 4.0 mmol/L
11.18
19.88
Any Lactate
Measurement
8.42
16.8
No Lactate
Measurement
5.85
13.94
All Sepsis Admissions
7.51
15.78
5.3.5
Change in Lactate
The ICU mortality rate of sepsis admissions, defined by infection and SIRS criteria,
with initial lactate greater than 4.0 mmol/L was much lower when the lactate decreased from the first to the second measurement than when the lactate increased, as
shown in figure 5-3. When lactate decreased by more than 80%, the ICU mortality
rate was only 5.00%. When lactate increased by more than 80%, the ICU mortality
rate was the highest at 71.43%. The ICU mortality rate had a large dip when lactate
59
Figure 5-3: Percent Change in Lactate From First to Second Measurement Vs. ICU
Mortality When the Initial Lactate Level Was Above 4.0 mmol/L for Sepsis Admissions Defined by Infection and SIRS Criteria
decreased by 20 to 40% due to the small sample size of patients with initial lactate
above 4.0 mmol/L whose lactate decreased by 20 to 40%. Overall, the ICU mortality
rate trend increased as the change in lactate increased from the first measurement to
the second measurement.
60
Chapter 6
Discussion
6.1
All ICU Admissions
We concluded that on average patients with high initial lactate levels were sicker
than those with low lactate levels. Patients with high initial lactate levels above
4.0 mmol/L had a much higher ICU mortality rate and 30 day mortality rate than
patients with normal lactate levels below 2.0 mmol/L. The ICU mortality rate and 30
day mortality rate for patients with high lactate levels were more than double the ICU
and 30 day mortality rates of patients with normal lactate levels. The average length
of stay of ICU survivors, a surrogate variable we used for how sick a patient was,
was a little longer for patients with high lactate than for patients with low lactate.
Admissions with high lactate typically were sicker, had longer ICU stays, and were
more likely to have poor outcomes.
To support our conclusions that patients with high initial lactate were sicker on
average than those with low initial lactate, we looked at average severity scores of
these populations. Admissions with low initial lactate had on average lower SOFA
scores and lower SAPS I scores than admissions with high initial lactate. A higher
severity score indicates that an illness is more severe and there is a greater likelihood of
a poor outcome. Patients with higher initial lactate had higher severity scores which
supported our original conclusion that in general patients with very high initial lactate
levels had worse outcomes and were sicker than patients with low initial lactate levels.
61
We also concluded that patients who do not get lactate levels measured are less
sick on average than patients who do get lactate levels measured. The ICU mortality
rate and 30 day mortality rate of patients without lactate measurements were less
than half of what the ICU mortality rate and 30 day mortality rate were for all ICU
admissions in our cohort. We also saw that patients without lactate measured had a
much shorter length of stay on average than patients with lactate measurements. In
general, we concluded that patients who were less sick in the ICU did not get a lactate
measurement, most likely because their physicians did not think the patients’ lactate
levels would be out of the normal range or need to be addressed. Those without
lactate measured who died in the ICU may have died quickly before any lactate could
be measured. Further studies would need to be done to confirm this hypothesis.
Looking at the admissions with lactate measurements in each care unit did not
give us much additional information. The ICU mortality rates of admissions with
lactate measurements in each care unit were slightly higher than the ICU mortality
rates of all patients in each care unit. The 30 day mortality rates in each care unit
followed the same trend. The same trend held for the average lengths of stay in each
care unit as well.
6.2
Septic Admissions
We concluded that admissions that were classified as septic or severely septic and that
had high initial lactate levels were typically sicker than admissions that were septic
with low initial lactate levels, the same trend we saw in all ICU patients. In all three
septic cohorts we studied, the ICU mortality rate and 30 day mortality rate increased
as the initial lactate level increased. The ICU mortality rate of admissions with
initial lactate levels above 4.0 mmol/L was more than double the ICU mortality rate
of admissions with initial lactate levels below 2.0 mmol/L in all three septic cohorts.
In addition, the 30 day mortality rates for admissions with high initial lactate were
almost twice as high as the 30 day mortality rates of admissions with low initial
lactate levels. Outcomes were especially poor for severe sepsis patients defined by the
62
Martin criteria; the 30 day mortality rate reached almost 60% for admissions with
high initial levels.
The average lengths of stay associated with each range of initial levels also showed
that on average septic patients with high lactate were sicker than septic patients with
low lactate. All of the septic cohorts showed the same trend; the higher the initial
lactate level, the longer the length of stay. A longer length of stay suggests that the
patients were sicker and needed attention for longer.
We investigated our conclusions about severity of illness in septic patients with
high lactate by looking at average SAPS I and SOFA scores. In all three septic
cohorts, we saw much higher average severity scores, indicating higher severity of
illness, in patients with high initial lactate compared to the average scores of patients
with low initial lactate. The average severity scores we found support our findings
that septic patients with high lactate are usually sicker than patients with low lactate.
6.3
Change in Lactate
Our results showed that change in lactate level is also very useful in predicting outcome, along with initial lactate level. We looked at change in lactate between the
first and second measurement in combination with the initial lactate level to gain
an even better understanding of the predictive value of lactate changes. We found
that looking at just lactate change by itself in all ICU patients did not show a large
difference in the ICU mortality rates. In all ICU admissions with two or more lactate
measurements, lactate levels that increased from the first to the second measurement
had a slightly higher mortality rate than admissions with lactate levels that decreased.
Looking at lactate change along with initial lactate level presented a much stronger
trend. ICU admissions that started with high lactate levels which increased had a
much higher mortality rate than those with high lactate levels that decreased.
In septic admissions, we saw an even greater effect of lactate change on ICU
mortality rates. There was a large increase in ICU mortality rates of septic admissions
when the initial lactate was high and the second measurement was higher than the
63
first. ICU mortality rates reached over 70% for all septic cohorts when initial lactate
was high and increased by a large amount, whereas ICU mortality rates were under
40% when the initial measurement was high but decreased. Knowing both the initial
lactate measurement and change in lactate has much more predictive value than either
measurement by itself, especially in septic patients.
It is still unknown why some admissions with high lactate that increases die while
others live. It is also unknown why some admissions with low initial lactate that
decreases die while others live. Further research is needed to understand the full
predictive value of initial lactate combined with lactate clearance.
During this study, we did not control for the time between the first and second
lactate measurements. There may be some admissions where the lactate was measured
very frequently, while others may have had days in between measurements. Looking
at the rate of change, as opposed to the just the percent change in lactate, may give
more information about how lactate change is associated with ICU mortality.
6.4
Comparisons Across Cohorts
Admissions within the septic cohorts typically had worse outcomes than admissions
in the total ICU cohort.The ICU mortality rates and 30 day mortality rates found for
each of the defined septic cohorts overall were higher than the ICU mortality rates
and 30 day mortality rates of the general ICU cohort. When including all patients
with or without a lactate measurement, Martin criteria severe septic admissions had
the highest ICU mortality rate overall, followed by the Angus criteria severe sepsis
admissions. Septic patients defined by infection and SIRS had a lower ICU mortality
rate than the other septic cohorts, but a slightly higher ICU mortality rate than all
ICU admissions. The same trend is seen in the 30 day mortality rates of each cohort.
The ICU mortality rate and 30 day mortality rate of admissions with high lactate
also vastly differed among the cohorts. The ICU mortality rate of Martin criteria
severe sepsis patients with high initial lactate (46.4%) was much higher than the ICU
mortality rates of patients with high initial lactate in the other cohorts. The next
64
highest ICU mortality rate associated with high initial lactate was 39.5%, which was
for patients in the Angus criteria defined cohort. After that, there was the infection
and SIRS defined septic cohort with an ICU mortality rate of 30.2%, and finally
after that was all ICU admissions with an ICU mortality rate of only 26.8%. The
30 day mortality rates follow the same decreasing trend from 59.2% (Martin criteria
admissions) to 41.6% (all ICU admissions). We concluded that patients identified
with the Martin criteria had the most severe illnesses, and that the Angus criteria
patients were a close second. It makes sense that these two cohorts were more ill than
the septic patients identified with infection and SIRS because Angus and Martin
criteria identify severe sepsis patients rather than just sepsis patients.
Although we saw large differences in mortality rates between the different cohorts,
we did not see large differences in severity scores. Septic patients had slightly higher
SOFA and SAPS I scores on average than all ICU admissions. All Martin criteria
patients with a lactate measured had an average SOFA score of 9.04 and an average
SAPS I score of 17.33, whereas the average SOFA and SAPS I scores of all ICU
admissions with a lactate level, 7.41 and 15.91 respectively, were only somewhat
lower. The average scores of the Angus criteria admissions and infection and SIRS
criteria admissions were in between as expected.
The average severity scores were much higher for patients with lactate above 4.0
mmol/L, but again we didn’t see a large difference across cohorts. The SOFA scores
ranged from 10.1 to 12.0 (all ICU admissions to Martin admissions) and the SAPS
I scores ranged from 19.2 to 20.8. The average severity scores of the three septic
cohorts differed by less than one point, when including the entire population and when
including just admissions with high initial lactate. Although we saw much different
ICU mortality rates and 30 day mortality rates across the cohorts, there were not large
differences in severity of illness based on severity scores. This shows that lactate levels
may give information about severity of illness in addition to the information given
by severity scores. Re-doing our methods but controlling for severity score may give
interesting results about how much additional information lactate levels can give us
about the severity of illness of a patient.
65
66
Chapter 7
Conclusion
7.1
Summary
In this thesis, we investigated the relationship between initial lactate levels and various easily measured parameters in ICU patients using the MIMIC II (version 2.6)
database. We found that increased lactate levels were associated with a higher ICU
mortality, higher 30 day mortality, longer ICU length of stay, and higher severity
scores.
In addition to looking at the ICU population as a whole, we identified sepsis
patients using three different criteria (Martin, Angus, and infection and SIRS) and
studied their lactate levels. We found higher ICU mortality rates, higher 30 day
mortality rates, longer lengths of stay, and higher severity scores in sepsis patients
compared to those of all ICU patients. Sepsis patients with high initial lactate levels
were the most severely ill of all the patients populations we investigated.
Of the three sepsis cohorts, the patients identified by the Martin criteria who
had high lactate levels had the worst outcomes. The patients identified by the Angus
criteria with high lactate had worse outcomes than the patients identified by infection
and SIRS criteria who had high lactate. Although we saw very large differences in
mortality rates, we did not see large differences in severity scores across the three
cohorts. This suggests that knowing initial lactate levels may add predictive value
in addition to severity scores. Lactate information may be important in addition to
67
classical severity scores in defining cohorts of patients.
We also investigated the relationship between initial lactate, change in lactate
from the first measurement to the second measurement, and ICU mortality. We
found that patients with high initial lactate levels in combination with an increase
in lactate level typically had poorer outcomes than patients with high initial lactate
levels with a decrease in lactate level.
7.2
Future Research Directions
There are a few different ways to further extend the research described in this thesis.
One interesting possibility would be to explore the role of lactate in additional subpopulations of ICU patients. In addition to sepsis patients, we found that patients
with cardiac problems and respiratory problems were two groups of patients that
might be at high risk for elevated lactate levels. In the future, we could define new
cohorts consisting of patients with respiratory or cardiac conditions and follow the
same methods that were presented in this thesis in order to learn more about the role
lactate in sub-populations of ICU patients.
Another possibility for future research would be to look more closely at the initial
lactate measurements and change in lactate. Some patients start out with very high
lactate levels and their lactate levels increase, but they still survive their ICU stay
despite this. Looking at other common factors associated with these patients such as
age, gender, and components of different severity scores may give light to why some
survive while others do not. We saw very low mortality rates in patients with normal
lactate whose lactate decreases, but there are still some patients that die. Identifying why some live while others die would be another question to solve. Being able
to predict mortality using lactate measurements in combination with other features
would be very useful to ICU doctors.
Following our methods again, but controlling for severity scores, is another idea
for the future. Severity scores give a lot of information about how likely it is that a
patient will die. Lactate measurements are not included in calculating the SOFA and
68
SAPS I scores, but lactate may give additional information about what a patient’s
outcome may be. Controlling for these scores would give us the ability to quantify
how much predictive value knowing lactate measurements adds.
One final future research plan is to re-run our methods and analysis on a much
larger cohort of patients to see if we obtain the same results. Our lab has recently
gained access to a database of 2.6 million patients and is in the process of adding
their information to the MIMIC II database (version 3). The patients are from many
different hospitals, so analyzing the new data would eliminate any bias that may have
been introduced in our current cohort, which is comprised of patients from only one
hospital.
69
70
Appendix A
A.1
Antibiotics used in Sepsis Definition
The antibiotics used to signify infection in septic patients were: ampicillin-sulbactam,
amoxicillin-clavulanic, amikacin, augmentin suspension, azithromycin, aztreonam,
bactrim, cefepime, cefixime, cefotaxime, cefotetan, cefoxitin, cefpodoxime, ceftazidime,
cefuroxime, ciprofloxacin, clarithromycin, clindamycin, colistin, daptomycin, doxycycline, ertapenem, erythromycin, gatifloxacin, gentamicin, imipenem, levofloxacin,
linezolid, meropenem, metronidazole, minocycline, moxifloxacin, piperacillin, quinupristin, synercid, tetracycline, tigecycline, timentin, tobramycin, trimethoprim, unasyn,
and vancomycin.
71
72
Bibliography
[1] Andra Blomkalns. Lactate-a measure for sepsis and trauma. Emergency Medicine
Cardiac Research and Education Group, September 2006.
[2] Jeffrey A. Kraut and Nicolaos E. Madias. Lactic acidosis. The New England
Journal of Medicine, (371):2309–2319, December 2014.
[3] Hitchcock R Trooskin SZ Henry SM Greenspan J Abramson D, Scalea TM.
Lactate clearance and survival following injury. J Trauma, 35(4):584–589, 1993.
[4] Talmor D et al. Shapiro NI, Howell MD. Serum lactate as a predictor of mortality
in emergency department patients with infection. Ann Emerg Med, 45(5):524–
528, 2005.
[5] Derek Angus et al. Epidemiology of severe sepsis in the united states: Analysis
of incidence, outcome, and associated costs of care. Critical Care Medicine,
29(7):1303–1310, July 2001.
[6] Mitchell M. Levy et al. 2001 sccm/esicm/accp/ats/sis international sepsis definitions conference. Intensive Care Medicine, 29(4):530–538, April 2003.
[7] Derek Angus and Tom van der Poll. Severe sepsis and septic shock. The New
England Journal of Medicine, (369):840–851, August 2013.
[8] Poeze Solberg and Ramsay Greve. Monitoring global volume-related hemodynamic or regional variables after initial resuscitation: What is a better predictor
of outcome in critically ill septic patients? Critical Care Medicine, 33(11):2494–
2500, 2005.
[9] Mohammed Saeed et al. Multiparameter intelligent monitoring in intensive
care ii: A public-access intensive care unit database. Critical Care Medicine,
39(5):952–960, 2011.
[10] Jean-Roger le Gall et al. A simplified acute physiology score for icu patients.
Critical Care Medicine, 12(11):975–977, 1984.
[11] Keu Sung Lee et al. Consideration of additional factors in sequential organ failure
assessment score. Journal of Critical Care, 29:185.e9–185.e12, 2014.
73
[12] Center
for
Disease
Control
and
Prevention.
classification
of
diseases,
ninth
revision,
clinical
http://www.cdc.gov/nchs/icd/icd9cm.htm, 2013.
International
modification.
[13] Greg S. Martin et al. The epidemiology of sepsis in the united states from 1979
through 2000. The New England Journal of Medicine, 348:1546–1554, April 2003.
[14] Theodore Iwashyna et al. Identifying patients with severe sepsis using administrative claims: Patient-level validation of the angus implementation of the international consensus conference definition of severe sepsis. Medical Care, 52(6):39–
43, June 2014.
74